# AdaFuse: Adaptive Ensemble Decoding with Test-Time Scaling for LLMs

arXiv: https://arxiv.org/abs/2601.06022 · HTML: https://mlsanity.com/paper/2601.06022/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 31 of its 39 references that are in the map.
- First posted on arXiv 9 Jan 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 30 of 31)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [Mistral 7B](https://mlsanity.com/paper/2310.06825/) (2023)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance](https://mlsanity.com/paper/2305.05176/) (2023)
- [LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion](https://mlsanity.com/paper/2306.02561/) (2023)
- [InternLM2 Technical Report](https://mlsanity.com/paper/2403.17297/) (2024)
- [RouterBench: A Benchmark for Multi-LLM Routing System](https://mlsanity.com/paper/2403.12031/) (2024)
- [Large Language Model Routing with Benchmark Datasets](https://mlsanity.com/paper/2309.15789/) (2023)
- [Routing to the Expert: Efficient Reward-guided Ensemble of Large Language Models](https://mlsanity.com/paper/2311.08692/) (2023)
- [Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory](https://mlsanity.com/paper/2511.20857/) (2025)
- [Fusing Models with Complementary Expertise](https://mlsanity.com/paper/2310.01542/) (2023)
- [Breaking the Ceiling of the LLM Community by Treating Token Generation as a Classification for Ensembling](https://mlsanity.com/paper/2406.12585/) (2024)
- [Determine-Then-Ensemble: Necessity of Top-k Union for Large Language Model Ensembling](https://mlsanity.com/paper/2410.03777/) (2024)
- [Improving LLM General Preference Alignment via Optimistic Online Mirror Descent](https://mlsanity.com/paper/2502.16852/) (2025)
- [Exact Byte-Level Probabilities from Tokenized Language Models for FIM-Tasks and Model Ensembles](https://mlsanity.com/paper/2410.09303/) (2024)
- [Cool-Fusion: Fuse Large Language Models without Training](https://mlsanity.com/paper/2407.19807/) (2024)
- [WAPITI: A Watermark for Finetuned Open-Source LLMs](https://mlsanity.com/paper/2410.06467/) (2024)
- [ResMoE: Space-efficient Compression of Mixture of Experts LLMs via Residual Restoration](https://mlsanity.com/paper/2503.06881/) (2025)
- [Life-Cycle Routing Vulnerabilities of LLM Router](https://mlsanity.com/paper/2503.08704/) (2025)
- [Harnessing Consistency for Robust Test-Time LLM Ensemble](https://mlsanity.com/paper/2510.13855/) (2025)
- [Improving Self Consistency in LLMs through Probabilistic Tokenization](https://mlsanity.com/paper/2407.03678/) (2024)
- [Hit the Sweet Spot! Span-Level Ensemble for Large Language Models](https://mlsanity.com/paper/2409.18583/) (2024)
- [RLAE: Reinforcement Learning-Assisted Ensemble for LLMs](https://mlsanity.com/paper/2506.00439/) (2025)

## Built on by (most cited first, 4 of 4)

- [Prune as You Generate: Online Rollout Pruning for Faster and Better RLVR](https://mlsanity.com/paper/2603.24840/) (2026)
- [Subspace Alignment for Vision-Language Model Test-time Adaptation](https://mlsanity.com/paper/2601.08139/) (2026)
- [Graph homophily booster: Reimagining the role of discrete features in heterophilic graph learning](https://mlsanity.com/paper/2602.07256/) (2026)
- [ReMix: Reinforcement routing for mixtures of LoRAs in LLM finetuning](https://mlsanity.com/paper/2603.10160/) (2026)

## Abstract (arXiv)

Large language models (LLMs) exhibit complementary strengths arising from differences in pretraining data, model architectures, and decoding behaviors. Inference-time ensembling provides a practical way to combine these capabilities without retraining. However, existing ensemble approaches suffer from fundamental limitations. Most rely on fixed fusion granularity, which lacks the flexibility required for mid-generation adaptation and fails to adapt to different generation characteristics across tasks. To address these challenges, we propose AdaFuse, an adaptive ensemble decoding framework that dynamically selects semantically appropriate fusion units during generation. Rather than committing to a fixed granularity, AdaFuse adjusts fusion behavior on the fly based on the decoding context, with words serving as basic building blocks for alignment. To be specific, we introduce an uncertainty-based criterion to decide whether to apply ensembling at each decoding step. Under confident decoding states, the model continues generation directly. In less certain states, AdaFuse invokes a diversity-aware scaling strategy to explore alternative candidate continuations and inform ensemble decisions. This design establishes a synergistic interaction between adaptive ensembling and test-time scaling, where ensemble decisions guide targeted exploration, and the resulting diversity in turn strengthens ensemble quality. Experiments on open-domain question answering, arithmetic reasoning, and machine translation demonstrate that AdaFuse consistently outperforms strong ensemble baselines, achieving an average relative improvement of 6.88%. The code is available at https://github.com/CCM0111/AdaFuse.